Learning useful representations without supervision remains a key challenge in machine learning. In this paper, we propose a simple yet powerful generative model that learns such discrete representations. Our model, the Vector Quantised-Variational AutoEncoder (VQ-VAE), differs from VAEs in two key ways: the encoder network outputs discrete, rather than continuous, codes; and the prior is learnt rather than static. In order to learn a discrete latent representation, we incorporate ideas from vector quantisation (VQ). Using the VQ method allows the model to circumvent issues of “posterior collapse” -— where the latents are ignored when they are paired with a powerful autoregressive decoder -— typically observed
in the VAE framework. Pairing these representations with an autoregressive prior, the model can generate high quality images, videos, and speech as well as doing high quality speaker conversion and unsupervised learning of phonemes, providing further evidence of the utility of the learnt representations.
Neural Discrete Representation Learning
最后編輯于 ：
?著作權歸作者所有,轉載或內容合作請聯系作者
- 文/潘曉璐 我一進店門员凝，熙熙樓的掌柜王于貴愁眉苦臉地迎上來，“玉大人奋献，你說我怎么就攤上這事健霹。” “怎么了瓶蚂？”我有些...
- 正文 為了忘掉前任钦听，我火速辦了婚禮洒试，結果婚禮上，老公的妹妹穿的比我還像新娘朴上。我一直安慰自己垒棋，他們只是感情好，可當我...
- 文/花漫 我一把揭開白布痪宰。 她就那樣靜靜地躺著叼架，像睡著了一般畔裕。 火紅的嫁衣襯著肌膚如雪。 梳的紋絲不亂的頭發(fā)上乖订，一...
- 文/蒼蘭香墨 我猛地睜開眼，長吁一口氣：“原來是場噩夢啊……” “哼眠饮！你這毒婦竟也來了奥帘？” 一聲冷哼從身側響起，我...
- 正文 年R本政府宣布嗦枢，位于F島的核電站，受9級特大地震影響屯断，放射性物質發(fā)生泄漏文虏。R本人自食惡果不足惜，卻給世界環(huán)境...
- 文/蒙蒙 一殖演、第九天 我趴在偏房一處隱蔽的房頂上張望氧秘。 院中可真熱鬧，春花似錦趴久、人聲如沸丸相。這莊子的主人今日做“春日...
- 文/蒼蘭香墨 我抬頭看了看天上的太陽灭忠。三九已至膳算，卻和暖如春，著一層夾襖步出監(jiān)牢的瞬間弛作，已是汗流浹背涕蜂。 一陣腳步聲響...
